Enhancement in Segmentation of Brain Tumors on MRI Using Region Growing Algorithm

In radiation oncology applications, the detection and segmentation of assorted necrotic and tumor tissue next to the adjacent vessels is a demanding situation. The MRI illustration produces an elevated fluctuate images indicative of usual in addition to unusual tissues that aid to discriminate the overlap into fringe of each tissue.

All habitual seed judgment methodologies might suffer through the difficulty but no development of tumor is there as well as if any small white part or grey part is present there. Segmentation of images by means of multifaceted structures such as magnetic resonance images of brain is tricky using common methods. The region based lively contour models are extensively used in segmentation of brain tumor.

If the boundaries of tumor is blunt, then the segmentation results are inaccurate i.e. segmentation could be above or beneath that may happen because of primary stage of the tumors. Here a scheme of tumor recognition based on texture of the MRI and the detected tumor is then segmented automatically using automatic seeded region growing technique and is proposed to detach the asymmetrical from the regular neighboring tissue to obtain an authentic detection of concerned and non-concerned area that facilitate the surgeon to discriminate the affected area accurately.

The methods used in this paper are texture analysis and automatic seeded region growing technique and is implemented on MRI of brain to identify the tumor margins in 2D MRI for dissimilar cases.